The complexity of modern microservice architectures has surpassed the capabilities of traditional orchestration tools, which rely on static, manual workflows. This limits scalability and adaptability, especially to dynamic workloads. This paper argues for a paradigm shift towards self-managing, intent-driven microservices orchestration systems, where human operators express high-level goals in natural language. As a foundational step towards this vision of autonomous orchestration agents, we introduce LEMON: a new architecture that leverages Large Language Models (LLMs) for intelligent microservice monitoring. We also propose a comprehensive classification to structure this emerging field. Our evaluation demonstrates that fine-tuning Small Language Models (SLMs) for intent classification significantly enhances accuracy while ensuring the model's output is reliably structured for automation. Furthermore, our analysis of the trade-offs between model size, precision, and latency provides a practical guide for deploying these systems. We foresee this monitoring capability as the critical ”sense” component toward autonomous microservices orchestration loops. By diagnosing performance bottlenecks from natural language queries, LEMON enables future systems to automatically suggest and execute solutions. This work lays the groundwork for truly self-managing, intent-driven systems.

Sparse linear algebra is a cornerstone of many scientific computing and machine learning applications. Python has become a popular choice for these applications due to its simplicity and ease of use. Yet high-performance sparse kernels in Python remain limited in functionality, especially on modern CPU and GPU architectures. We present pyGinkgo, a lightweight and Pythonic interface to the Ginkgo library, offering high-performance sparse linear algebra support with platform portability across CUDA, HIP, and OpenMP backends. pyGinkgo bridges the gap between high-performance C++ backends and Python usability by exposing Ginkgo’s capabilities via Pybind11 and a NumPy and PyTorch compatible interface. We benchmark pyGinkgo’s performance against state-of-the-art Python libraries including SciPy, CuPy, PyTorch and TensorFlow. Results across hardware from different vendors demonstrate that pyGinkgo consistently outperforms existing Python tools in both Sparse Matrix Vector (SpMV) product and iterative solver performance, while maintaining performance parity with native Ginkgo C++ code. Our work positions pyGinkgo as a compelling backend for sparse machine learning models and scientific workflows.

The gradients used to train neural networks are typically computed using backpropagation. While an efficient way to obtain exact gradients, backpropagation is computationally expensive, hinders parallelization, and is biologically implausible. Forward gradients are an approach to approximate the gradients from directional derivatives along random tangents computed by forward-mode automatic differentiation. So far, research has focused on using a single tangent per step. This paper provides an in-depth analysis of multi-tangent forward gradients and introduces an improved approach to combining the forward gradients from multiple tangents based on orthogonal projections. We demonstrate that increasing the number of tangents improves both approximation quality and optimization performance across various tasks.

This study explores the learning dynamics of neural networks by analyzing the singular value decomposition (SVD) of their weights throughout training. Our investigation reveals that an orthogonal basis within each multidimensional weight’s SVD representation stabilizes during training. Building upon this, we introduce Orthogonality-Informed Adaptive Low-Rank (OIALR) training, a novel training method exploiting the intrinsic orthogonality of neural networks. OIALR seamlessly integrates into existing training workflows with minimal accuracy loss, as demonstrated by benchmarking on various datasets and well-established network architectures. With appropriate hyperparameter tuning, OIALR can surpass conventional training setups, including those of state-of-the-art models.

Deep learning has become a popular tool for solving complex problems in a variety of domains. Transformers and the attention mechanism have contributed a lot to this success. We hypothesize that the enhanced predictive capabilities of the attention mechanism can be attributed to higher-order terms in the input. Expanding on this idea and taking inspiration from Taylor Series approximation, we introduce “Taylor layers” as higher order polynomial layers for universal function approximation. We evaluate Taylor layers of second and third order on the task of time series forecasting, comparing them to classical linear layers as well as the attention mechanism. Our results on two commonly used datasets demonstrate that higher expansion orders can improve prediction accuracy given the same amount of trainable model weights. Interpreting higher-order terms as a form of token mixing, we further show that second order (quadratic) Taylor layers can efficiently replace canonical dot-product attention, increasing prediction accuracy while reducing computational requirements.

In the last decade, deep learning (DL) has significantly impacted industry and science. Initially largely motivated by computer vision tasks in 2-D imagery, the focus has shifted toward 3-D data analysis. In particular, 3-D surface reconstruction, i.e., reconstructing a 3-D shape from sparse input, is of great interest to a large variety of application fields. DL-based approaches show promising quantitative and qualitative surface reconstruction performance compared to traditional computer vision and geometric algorithms. This survey provides a comprehensive overview of these DL-based methods for 3-D surface reconstruction. To this end, we will first discuss input data modalities, such as volumetric data, point clouds, and RGB, single-view, multiview, and depth images, along with corresponding acquisition technologies and common benchmark datasets. For practical purposes, we also discuss evaluation metrics enabling us to judge the reconstructive performance of different methods. The main part of the document will introduce a methodological taxonomy ranging from point-and mesh-based techniques to volumetric and implicit neural approaches. Recent research trends, both methodological and for applications, are highlighted, pointing toward future developments.

BigEarthNet is one of the standard large remote sensing datasets. It has been shown previously that neural networks are effective tools to classify the image patches in this data. However, finding the optimum network hyperparameters and architecture to accurately classify the image patches in BigEarthNet remains a challenge. Searching for more accurate models manually is extremely time consuming and labour intensive. Hence, a systematic approach is advisable. One possibility is automated evolutionary Neural Architecture Search (NAS). With this NAS many of the commonly used network hyperparameters, such as loss functions, are eliminated and a more accurate network is determined.

To cope with the rapid growth in available data, the efficiency of data analysis and machine learning libraries has recently received increased attention. Although great advancements have been made in traditional array-based computations, most are limited by the resources available on a single computation node. Consequently, novel approaches must be made to exploit distributed resources, e.g. distributed memory architectures. To this end, we introduce HeAT, an array-based numerical programming framework for large-scale parallel processing with an easy-to-use NumPy-like API. HeAT utilizes PyTorch as a node-local eager execution engine and distributes the workload on arbitrarily large high-performance computing systems via MPI. It provides both low-level array computations, as well as assorted higher-level algorithms. With HeAT, it is possible for a NumPy user to take full advantage of their available resources, significantly lowering the barrier to distributed data analysis. When compared to similar frameworks, HeAT achieves speedups of up to two orders of magnitude.

When training a classifier the choice of loss function heavily influences the characteristics of the resulting model. The most commonly used loss function for classification is cross entropy. In image segmentation problems where each pixel is assigned to a particular class, overlap-based losses have recently been shown to improve classifier performance especially for datasets with an imbalanced class distribution. This is particu-larly relevant to segmentation because class imbalance mitigation strategies used in regular classification are often not applicable. Overlap-based losses, however, have different drawbacks. We are aiming at combining the upsides of different losses with a simple scheduling scheme during training while minimizing their downsides. Gradually transitioning from an overlap-based dice loss to cross entropy allows to reliably select a distinct minimum in the optimization landscape as a valuable alternative to results obtained from traditional unscheduled loss functions. We demonstrate the efficacy of our approach on different combinations of loss functions, datasets, and models.

Supervised image classification is one of the essential techniques for generating semantic maps from remotely sensed images. The lack of labeled ground truth datasets, due to the inherent time effort and cost involved in collecting training samples, has led to the practice of training and validating new classifiers within a single image. In line with that, the dominant approach for the division of the available ground truth into disjoint training and test sets is random sampling. This paper discusses the problems that arise when this strategy is adopted in conjunction with spectral-spatial and pixel-wise classifiers such as 3D Convolutional Neural Networks (3D CNN). It is shown that a random sampling scheme leads to a violation of the independence assumption and to the illusion that global knowledge is extracted from the training set. To tackle this issue, two improved sampling strategies based on the Density-Based Clustering Algorithm (DBSCAN) are proposed. They minimize the violation of the train and test samples independence assumption and thus ensure an honest estimation of the generalization capabilities of the classifier.

Component trees are region-based representations that encode the inclusion relationship of the threshold sets of an image. These representations are one of the most promising strategies for the analysis and the interpretation of spatial information of complex scenes as they allow the simple and efficient implementation of connected filters. This work proposes a new efficient hybrid algorithm for the parallel computation of two particular component trees-the max- and min-tree-in shared and distributed memory environments. For the node-local computation a modified version of the flooding-based algorithm of Salembier is employed. A novel tuple-based merging scheme allows to merge the acquired partial images into a globally correct view. Using the proposed approach a speed-up of up to 44.88 using 128 processing cores on eight-bit gray-scale images could be achieved. This is more than a five-fold increase over the state-of-the-art shared-memory algorithm, while also requiring only one-thirty-second of the memory.

In this paper, we propose an instrumentation and computer vision pipeline that allows automatic object detection on images taken from multiple experimental set ups. We demonstrate the approach by autonomously counting intoxicated flies in the FLORIDA assay. The assay measures the effect of ethanol exposure onto the ability of a vinegar fly Drosophila melanogaster to right itself. The analysis consists of a three-step approach. First, obtaining an image of a large set of individual experiments, second, identify areas containing a single experiment, and third, discover the searched objects within the experiment. For the analysis we facilitate well-known computer vision and machine learning algorithms - namely color segmentation, threshold imaging and DBSCAN. The automation of the experiment enables an unprecedented reproducibility and consistency, while significantly decreasing the manual labor.

Big Data Analytics methods take advantage of techniques from the fields of data mining, machine learning, or statistics with a focus on analysing large quantities of data (aka `big datasets') with modern technologies. Big data sets appear in remote sensing in the sense of large volumes, but also in the sense of an ever increasing amount of spectral bands (i.e., high-dimensional data). The remote sensing has traditionally used the above described techniques for a wide variety of application such as classification (e.g., land cover analysis using different spectral bands from satellite data), but more recently scalability challenges occur when using traditional (often serial) methods. This paper addresses observed scalability limits when using support vector machines (SVMs) for classification and discusses scalable and parallel developments used in concrete application areas of remote sensing. Different approaches that are based on massively parallel methods are discussed as well as recent developments in parallel methods.

The big data analytics approach emerged that can be interpreted as extracting information from large quantities of scientific data in a systematic way. In order to have a more concrete understanding of this term we refer to its refinement as smart data analytics in order to examine large quantities of scientific data to uncover hidden patterns, unknown correlations, or to extract information in cases where there is no exact formula (e.g. known physical laws). Our concrete big data problem is the classification of classes of land cover types in image-based datasets that have been created using remote sensing technologies, because the resolution can be high (i.e. large volumes) and there are various types such as panchromatic or different used bands like red, green, blue, and nearly infrared (i.e. large variety). We investigate various smart data analytics methods that take advantage of machine learning algorithms (i.e. support vector machines) and state-of-the-art parallelization approaches in order to overcome limitations of big data processing using non-scalable serial approaches.